FormalPara Key Summary Points

Respiratory syncytial virus (RSV) incidence is usually underestimated among adults.

We estimated RSV-attributable hospitalizations and deaths among adults in Italy during 2015–2019 using a quasi-Poisson regression model.

Estimated RSV hospitalization incidence rates were substantially higher than those based on RSV-specific ICD codes only.

RSV contributes to both respiratory and cardiovascular disease in adults.

RSV-attributable hospitalization and mortality incidences were highest among adults aged ≥ 75 years.

Introduction

RSV is a common cause of respiratory infections in adults, especially at older age. A recent global systematic literature review and meta-analysis of incidence among older adults from high-income countries found that the annual pooled incidence rate (IR) of RSV-related acute respiratory infection hospitalizations adjusted for case under-ascertainment was 347 per 100,000 population [1]. RSV infections can also trigger acute cardiovascular events, such as congestive heart failure, ischemic events, arrhythmias, and stroke [2, 3]. Individuals with comorbidities such as cardiorespiratory comorbidities or immunocompromising conditions are at a higher risk of developing severe RSV disease and have poorer outcomes [4,5,6,7]. In Europe, RSV shows a seasonal pattern, with epidemics occurring in the winter months [8]. The start of seasonal RSV epidemics is often characterized by an increase in the incidence of acute bronchiolitis in children [9].

New vaccines against RSV were recently licensed to prevent lower respiratory tract disease in older adults aged 60 years and above [10, 11]. To assess potential vaccine impact, a more accurate estimation of RSV incidence and burden in adults is essential. The primary constraint on more accurate disease burden estimates is RSV’s non-specific symptomology compared to influenza or other respiratory viruses [12]. In addition, RSV episodes are underestimated due to limited standard-of-care testing among adults and lower sensitivity of testing compared to infants [13,14,15,16,17].

To address these limitations, much of the current RSV burden evidence has come from modeling studies [18,19,20,21,22,23]. These model-based approaches retrospectively estimate RSV incidence using electronic medical records or health insurance claims. The models estimate the proportion of health outcomes attributable to RSV by correlating variations in non-specific health outcomes (e.g., cardiorespiratory diseases, pneumonia, exacerbation of chronic respiratory diseases) with variations in RSV activity while adjusting for the activity of other viral infections, seasonality, and underlying temporal trends. The incidence estimates derived from these models closely align with prospective study estimates, outperforming those based on standard-of-care testing and RSV-specific ICD-10 codes [24].

Some studies have partially assessed the epidemiology of RSV in Italian adults; however, these studies either did not provide adult-specific incidence estimates or were conducted during the COVID-19 pandemic when RSV epidemiology was distorted due to nonpharmaceutical interventions [25, 26]. Therefore, in this study, we estimated the epidemiologic burden of RSV disease in the Italian adult population before the COVID-19 pandemic between 2015 and 2019.

Methods

Study Design

We performed a retrospective analysis of national hospitalization and mortality databases to estimate the incidences of RSV-attributable hospitalizations and mortality. This involved employing a statistical modeling approach that links the variation of RSV activity to the variation of selected hospitalizations or deaths while adjusting for influenza activity, seasonality, and underlying temporal trends. Influenza is an important co-circulating pathogen that is controlled for while estimating the burden of RSV. However, as our main objective was to estimate RSV burden, the results presented in this paper focus solely on RSV. We published the detailed generic study protocol that was implemented in this study [28].

Data Sources

In absence of weekly aggregated data, monthly hospitalization data were requested and produced by the Italian Center for Applied Economic Research in Healthcare (CREA Sanità) by extracting data from the Ministry of Health database. Similarly, due to unavailability of weekly data, monthly data on deaths (2015–2019) were obtained from the national database of the Italian National Institute of Statistics (ISTAT). Hospitalizations were coded according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), the classification currently used in the Italian health database, and mortality data according to ICD-10-CM. We selected two primary outcomes for modeling: all cardiorespiratory diseases and all respiratory diseases. As in other model-based studies [29,30,31,32], we investigated the following subgroups of cardiorespiratory diseases: influenza or pneumonia, chronic lower respiratory diseases, chronic heart failure exacerbation, and ischemic heart diseases, which were found to be related to RSV [3, 6, 33].

The outcome definitions are provided in the Supplementary Table 1. We considered primary and secondary diagnoses, as reported in hospital discharge records, and the underlying cause of death in death certificates.

Individuals were categorized into four age groups: 18–44, 45–64, 65–74, and ≥ 75 years. Hospitalization data were stratified by RSV risk factors, defined as the presence of at least one comorbidity reported within 1 year before the event. Chronic conditions, including cardiac, respiratory, liver and kidney disease, immunosuppression, neurological disorders, and diabetes mellitus, were considered risk factors [6]. As the literature on risk factors for RSV among adults is limited [5], we also considered risk factors for severe influenza infection [34]. The list of comorbidities is provided in Supplementary Table 2.

We obtained counts of age-specific populations at risk of the event (denominator data) from national census data from ISTAT.

We defined the RSV proxy as the number of monthly RSV-specific hospitalizations (corresponding to ICD-9 codes: 079.6, 466.1, 466.11, 480.1) in children < 2 years, and the influenza proxy as the number of influenza-specific hospitalizations (corresponding to ICD-9-CM codes: 487 and 488) in people ≥ 65 years. Similar to other studies, we used pediatric RSV activity for the RSV proxy because testing is much more frequent among young children, allowing for consistent measurement of RSV activity [18, 35, 36]. In addition, the temporal association between RSV waves in children and adults (after different lags), has been demonstrated previously [37, 38]. We included acute bronchiolitis (466.1), because RSV is responsible for the majority of bronchiolitis hospitalizations in this age group [39,40,41]. The latter also serves as an indicator marking the start of the RSV season and tracks RSV activity independently of RSV testing levels. We used influenza-specific hospitalizations in older adults for the influenza activity proxy, as the largest burden of influenza and most consistent testing is among older adults [22].

Statistical Analysis

As described in the published generic study protocol [28], we used a quasi-Poisson regression model to model the number of events for each subgroup [i.e., health outcome per age group and risk group (only for hospitalizations)] as a function of periodic time trends, aperiodic time trends, and viral activity (RSV and influenza) while allowing for potential overdispersion. We constructed the general models for hospitalizations and mortality as described in the published time-series model-based analysis protocol to estimate RSV incidence rate in adults [28]. To allow for the additional flexibility required to adequately capture the drops in the number of hospitalizations recurring yearly in December, we included both yearly and half-yearly harmonics in the model for hospitalizations. Since RSV and influenza have similar seasonal patterns and both contribute to the outcomes of interest, they were included in the model as covariates. In addition, an overdispersion term was added to account for other minor respiratory such as rhinovirus pathogens.

From the subgroup-specific models, we obtained the monthly number of RSV-attributable events by calculating the difference between the expected number of events from the full model and that from the model without the RSV term (by setting the coefficient associated with RSV to zero). We obtained the yearly number of RSV-attributable events, stratified by age group, and, when applicable, risk group, by summing the study years.

We obtained IR estimates stratified by year and age group by dividing the annual model-based number of events by the corresponding number of individuals at risk of the event (denominator data). As denominator data were not available at the risk group level, we did not calculate the risk group-specific IRs. The age-specific number of events was obtained by summing the model-based number of events over the risk groups.

To compare the difference between the observed (reported in the database regarding standard-of-care testing) versus the attributable (model-based) IR of RSV events, we also calculated the annual IR of RSV-specific hospitalizations (any of the following ICD-9-CM codes: 079.6, 466.11, 480.1) per age group.

We performed data management and statistical analysis in R version 4.2.2.

Ethical Considerations

We conducted the study following legal and regulatory requirements, with a scientific purpose, value and rigor. We followed generally accepted research practices described in the Good Epidemiological Practice guidelines issued by the International Epidemiological Association [42].

This study involved anonymized structured data without patients' personal information and required no ethical approval.

Results

Recorded Events

In the period 2015–2019, a total of, respectively, 10,931,573 and 1,390,267 cardiorespiratory hospitalizations and deaths were recorded among adults in Italy. Individuals aged ≥ 65 years accounted for 74% and 94% of all cardiorespiratory hospitalizations and deaths, respectively.

The IR of reported RSV-specific hospitalizations based on RSV-specific ICD coding alone ranged from 0.1 to 3.7 cases per 100,000 population (Table 1).

Table 1 Recorded annual RSV-specific hospitalization incidence rates per 100,000 population by age group per RSV diagnoses recorded with RSV-specific ICD codes, 2015–2019, Italy
Full size table

Estimated RSV-Attributable Hospitalizations

A seasonal pattern of hospitalizations was observed for the six health outcomes and all age groups, except for 18–44 years, with persistent drops in reported episodes in December of each year. The model fitted relatively well with the observed data in people with and without risk factors, with some outliers, as shown in Supplementary Figs. 1 and 2. For most outcomes, the age group 18–44 years were not modeled because the observed data did not show a clear seasonal pattern.

During the study period, a total of 879,351 cardiorespiratory hospitalizations were attributable to RSV in persons aged ≥ 45 years. The estimated number and annual IRs of RSV-attributable hospitalizations stratified by outcome and age group are presented in Table 2. The IRs were highest for RSV-attributable cardiorespiratory hospitalizations, followed by respiratory and chronic heart failure exacerbation hospitalizations. Regardless of the age group and the outcome, the IRs increased over the studied years. In all outcomes, IRs increased with age. Overall, the estimates in the age group ≥ 75 years were approximately twofold higher than that at 65–74 years. Incidence rates for RSV-attributable hospitalizations were on average 2–3 times higher for cardiorespiratory than respiratory disease alone.

Table 2 Estimated annual RSV-attributable hospitalization incidence rates per 100,000 population, by diagnosis and by age group, 2015–2019, Italy
Full size table

The estimated number of hospitalizations varied by age group and study outcomes: between 45 and 64 years, it was higher in people without risk factors for all cardiorespiratory and respiratory diseases, while the contrary was found ≥ 65 years (Supplementary Table 3). For the subgroups of cardiorespiratory diseases, the estimated number of hospitalizations was higher in those with risk factors, especially for those with chronic heart failure. The proportion of hospitalizations attributable to RSV was comparable between people with and without risk factors, but a higher proportion was observed in the younger age groups.

Estimated RSV-Attributable Mortality

A seasonal pattern was observed for the six mortality outcomes in the age groups of 45 years and above. Data did not allow modeling in the 18–44 years age group, as there was a low number of deaths (< 2000 cardiorespiratory deaths per year), and no clear seasonal pattern was observed in this age group. There was a good model fit for the remaining age groups for all outcomes except for chronic heart failure exacerbation and ischemic heart diseases (Supplementary Figs. 3 and 4). For the latter, we either did not obtain a good model fit (45–64 years) or obtained unstable results (negative estimates) (age groups ≥ 65 years).

A total of 26,824 cardiorespiratory deaths were attributed to RSV among people aged ≥ 45 years between 2015 and 2019. The estimated number of deaths and mortality rates (MRs) attributed to RSV are presented in Table 3. The MRs increased consistently over the studied years for all outcomes and age groups. In people aged ≥ 65 years, the highest MRs were found for cardiorespiratory diseases, but RSV contributed to a larger proportion of deaths for respiratory diseases (3.5–4.7%) than for cardiorespiratory diseases (1.6–2.5%). The number of RSV-attributable deaths and MRs in the age group 45–64 years was minimal for all outcomes.

Table 3 Estimated RSV-attributable mortality rate per 100,000, by diagnosis and by age group, 2015–2019, Italy
Full size table

The MR of RSV-attributable cardiorespiratory deaths increased with age for all outcomes; it was approximately 13-fold higher among those aged ≥ 75 years compared to that in the 65–74 years age group.

Discussion

According to our model, RSV was deemed to be responsible for a considerable cardiorespiratory disease burden among adults in Italy during the period 2015–2019, which was 405- to 1729-fold higher than estimates based on RSV-specific ICD codes alone. We accounted for the possibility of RSV to trigger acute cardiac events, and including cardiovascular disease as well as respiratory illness increased our incidence/mortality rate estimates by two- to threefold. As expected, we found that RSV-attributable incidences were especially high among older adults (aged ≥ 65 years), who accounted for 78% of RSV-attributable hospitalizations.

There was a sharp rise in disease burden between the two oldest age groups with the rates for adults ≥ 75 years were up to 2-fold (for hospitalizations) and 13-fold (for mortality) higher than those in adults 65–74 years. These results are in line with other studies, which indicate that RSV disease burden typically follows a U-shaped age pattern [4, 43, 44], with the highest hospitalization and mortality rates found among infants and older adults The positive association between age and infectious disease in adults is well known and related to several factors, such as the natural decline of the immune response with age (i.e., immunosenescence), as well as the high proportion of older adults with underlying diseases [4, 45]. Consistent with this, we observed that patients with risk factors represented approximately half of cardiorespiratory hospitalizations among older adults [vs. ~ 19% among 18–44 years (data not presented)].

RSV contributed not only to respiratory but also to cardiovascular hospitalizations. For instance, RSV was accountable for 11% of chronic heart failure exacerbation hospitalizations on average in adults aged ≥ 45 years. RSV is able to exacerbate pre-existing cardiovascular diseases like coronary artery disease or heart failure, trigger new cardiovascular events, and increase the risk of subsequent cardiovascular hospitalizations post-hospitalization for lower respiratory infection [2, 3]. Part of these clinical manifestations are due to indirect effects originating from the infected respiratory tract (e.g., due to pulmonary hypertension secondary to severe RSV bronchitis), acute hypoxia, or from the inflammatory response. However, the direct involvement of RSV in myocardial injury has also been suggested [3, 46]. Overall, although RSV contributed to similar proportions of respiratory and cardiorespiratory hospitalizations, the attributable numbers and incidences were higher for cardiorespiratory causes because of the high number of all-cause cardiovascular hospitalizations. While most RSV burden studies focus solely on respiratory disease [1, 24], adding cardiovascular disease increased incidence and mortality rate estimates by 2–3 times.

A study conducted between 2007 and 2021 in Italy in the Northern Veneto region [47] estimated an annual acute lower respiratory infection hospitalization incidence of 0.1–3.8 per 100,000 in subjects aged ≥ 65 years, based on RSV-specific ICD codes. These results are in line with our observed/unmodeled RSV hospitalization rates based on RSV-specific ICD codes (0.6–2.8 per 100,000; Table 1). Such analyses based on ICD codes diagnosis alone have been shown to substantially underestimate RSV burden, largely due to infrequent RSV testing in adults as well as RSV infection being coded into other non-pathogen-specific conditions such as pneumonia [15, 48]. The substantial difference between the latter and our model-based estimated RSV-attributable hospitalization rates for all respiratory diseases (262–381 per 100,000 population) again highlights that a significant proportion of RSV disease burden is not apparent. For instance, our model estimated that RSV was responsible for 5.7–8.5% and 5.6–7.1% of the respiratory and the pneumonia or influenza hospitalizations among adults ≥ 65 years, respectively. In line with these findings, a molecular surveillance study of RSV infection conducted in Italy (Sicily) found an RSV-positivity rate of 5.1% among patients hospitalized for influenza-like illness or severe respiratory infections during the five winter seasons (October–April) preceding the onset of the COVID-19 pandemic [49]. In another prospective multicenter study conducted in Italy during the 2018–2019 among patients aged 65–80 and ≥ 80 years, 10% and 12%, respectively, of samples were positive to RSV [44]. The observed high positivity rate in this study might be related to the restriction of the study to only one epidemic season and different criteria used for testing.

Similar to the Veneto region’s study [47], we observed an increase in rates over the study period. This positive trend across the years is most likely related to increase in the measured RSV activity in the most recent years due to improved diagnostic procedures [47, 50]. At the same time, the number of cardiorespiratory hospitalizations and deaths remained stable. Our IRs for respiratory hospitalizations in individuals aged 45–64 years and ≥ 65 years (81.2–117.3 and 262–381 per 100,000, respectively) are consistent with recent pooled estimates from high-income countries, adjusted for under-ascertainment which reported 150 hospitalizations per 100,000 in persons aged 50–64 years [51], and 236 [24] and 347 [1] hospitalizations per 100,000 in older adults. Our results are also very similar to those obtained in Spain and Germany using the same modeling method as in our study, which reported, respectively, 261–283 and 236–363 respiratory hospitalizations per 100,000 person-years in adults aged ≥ 60 years [52, 53].

Pre-existing comorbidities have been shown to be associated with more severe RSV disease, increased hospitalization need and duration, and higher rates of intensive care unit admission and mortality [6]. Unfortunately, because denominator data were not available, we were not able to verify this association by stratifying the RSV-attributable hospitalization rates by risk factors. Nevertheless, we observed for all outcomes that the numbers of RSV-attributable hospitalizations among older adults were higher in patients with risk factors than in those without (Supplementary Table 3). We also found that, except for chronic lower respiratory diseases, hospitalization counts attributable to RSV were globally higher among adults aged 45–64 years without than with risk factors, which is most likely related to the lower prevalence of risk factors in this age group. Because we constructed the risk groups by considering a look-back period limited to 1 year prior to hospital admission, some patients with risk factors might have been misclassified, which could have impacted our results.

In our study, a substantial burden of deaths could be attributed to RSV in older people, with the vast majority of those (93% of cardiorespiratory deaths) occurring among adults aged ≥ 75 years. The lower number of underlying cardiovascular deaths did not allow modeling for the youngest age group (18–44 years), which also explains the negative and zero estimates that we obtained for the age group 45–64 years. Moreover, the unstable modelling results obtained in the age groups ≥ 65 years for chronic heart failure exacerbation and ischemic heart diseases are most likely due to the combination of several factors: (1) less clear mortality peaks compared to other outcomes, (2) the overlap of the RSV and influenza proxies seasonality, and (3) the limited year-to-year variability. Our rates of underlying respiratory deaths in adults aged ≥ 65 years (12.1–17.6 per 100,000) were comparable to the results of a recently published model-based study conducted in the United States (14.7 deaths per 100,000 people aged ≥ 65 years per year) [38] and to results obtained in Spain using the same modeling method as in our study (14.6–17.1 deaths per 100,000 person-years in people aged ≥ 60 years) [52].

To our knowledge, this is the first model-based study that has estimated RSV-associated hospitalization and mortality rates among adults in Italy. One of the strengths of our study is the use of nationally representative hospitalization and mortality databases, which reduces the possibility of sampling bias. By taking into account an extensive list of ICD codes corresponding to broad definitions for cardiorespiratory and respiratory outcomes and screening the hospitalization data for both primary and secondary diagnoses, we could more reliably capture the burden of RSV disease. The quasi-Poisson regression model allowed us to account for overdispersion and to adjust for the co-circulation of influenza while correcting for underlying time trends.

Our study had several limitations. As weekly counts of health outcomes were not available, we had to rely on monthly counts, thereby limiting the number of time points included in our model and not allowing us to chose the exact time lag between viral activity and outcomes the way it could have been done with weekly data [28]. Nevertheless, our study fills a knowledge gap in the best achievable way given the available data in Italy; moreover, other modeling studies have estimated RSV disease burden relying on monthly data [36, 54, 55]. It should also be noted that the recurrent drop in all-cause hospitalizations in December every year, related to the Christmas holidays, could have impacted our model for the hospitalizations; nevertheless, the model fits were overall good (Supplementary Fig. 1), and the results of sensitivity analyses (data not presented), in which we imputed December counts, demonstrated that our model was robust. Also, as an RSV activity proxy, we used hospitalization data from children aged < 2 years (consistent with another publication [36], which may not reflect the exact temporality of circulation of the virus in the adult population, but is the best available proxy for RSV as explained above. We also did not include proxies for the activity of respiratory pathogens other than RSV and influenza in our model, which might have resulted in an overestimation of the RSV burden. However, even without explicitly modeling other potentially relevant pathogens, they are indirectly accounted for in the model through the periodic component and the overdispersion parameter. Lastly, we estimated the burden of RSV disease in the Italian adult population before the COVID-19 pandemic. However, it has been suggested that the epidemiology of RSV post-COVID pandemic closely resembles that before the pandemic [27]. In consequence, our pre-pandemic estimates still provide meaningful insight into the current RSV burden.

Conclusion

Based on our model results, we can conclude that RSV is responsible for a considerable and largely underestimated burden of hospitalizations and deaths among adults in Italy. Like other respiratory viruses, RSV contributes to both respiratory and cardiovascular conditions. In total, these data emphasize the need to implement effective RSV prevention strategies, particularly among the most vulnerable older adults.